Artificial Intelligence Nanodegree

Convolutional Neural Networks

Project: Write an Algorithm for a Dog Identification App


In this notebook, some template code has already been provided for you, and you will need to implement additional functionality to successfully complete this project. You will not need to modify the included code beyond what is requested. Sections that begin with '(IMPLEMENTATION)' in the header indicate that the following block of code will require additional functionality which you must provide. Instructions will be provided for each section, and the specifics of the implementation are marked in the code block with a 'TODO' statement. Please be sure to read the instructions carefully!

Note: Once you have completed all of the code implementations, you need to finalize your work by exporting the iPython Notebook as an HTML document. Before exporting the notebook to html, all of the code cells need to have been run so that reviewers can see the final implementation and output. You can then export the notebook by using the menu above and navigating to \n", "File -> Download as -> HTML (.html). Include the finished document along with this notebook as your submission.

In addition to implementing code, there will be questions that you must answer which relate to the project and your implementation. Each section where you will answer a question is preceded by a 'Question X' header. Carefully read each question and provide thorough answers in the following text boxes that begin with 'Answer:'. Your project submission will be evaluated based on your answers to each of the questions and the implementation you provide.

Note: Code and Markdown cells can be executed using the Shift + Enter keyboard shortcut. Markdown cells can be edited by double-clicking the cell to enter edit mode.

The rubric contains optional "Stand Out Suggestions" for enhancing the project beyond the minimum requirements. If you decide to pursue the "Stand Out Suggestions", you should include the code in this IPython notebook.


Why We're Here

In this notebook, you will make the first steps towards developing an algorithm that could be used as part of a mobile or web app. At the end of this project, your code will accept any user-supplied image as input. If a dog is detected in the image, it will provide an estimate of the dog's breed. If a human is detected, it will provide an estimate of the dog breed that is most resembling. The image below displays potential sample output of your finished project (... but we expect that each student's algorithm will behave differently!).

Sample Dog Output

In this real-world setting, you will need to piece together a series of models to perform different tasks; for instance, the algorithm that detects humans in an image will be different from the CNN that infers dog breed. There are many points of possible failure, and no perfect algorithm exists. Your imperfect solution will nonetheless create a fun user experience!

The Road Ahead

We break the notebook into separate steps. Feel free to use the links below to navigate the notebook.

  • Step 0: Import Datasets
  • Step 1: Detect Humans
  • Step 2: Detect Dogs
  • Step 3: Create a CNN to Classify Dog Breeds (from Scratch)
  • Step 4: Use a CNN to Classify Dog Breeds (using Transfer Learning)
  • Step 5: Create a CNN to Classify Dog Breeds (using Transfer Learning)
  • Step 6: Write your Algorithm
  • Step 7: Test Your Algorithm

Step 0: Import Datasets

Import Dog Dataset

In the code cell below, we import a dataset of dog images. We populate a few variables through the use of the load_files function from the scikit-learn library:

  • train_files, valid_files, test_files - numpy arrays containing file paths to images
  • train_targets, valid_targets, test_targets - numpy arrays containing onehot-encoded classification labels
  • dog_names - list of string-valued dog breed names for translating labels
In [1]:
from sklearn.datasets import load_files       
from keras.utils import np_utils
import numpy as np
from glob import glob

# define function to load train, test, and validation datasets
def load_dataset(path):
    data = load_files(path)
    dog_files = np.array(data['filenames'])
    dog_targets = np_utils.to_categorical(np.array(data['target']), 133)
    return dog_files, dog_targets

# load train, test, and validation datasets
train_files, train_targets = load_dataset('dogImages/train')
valid_files, valid_targets = load_dataset('dogImages/valid')
test_files, test_targets = load_dataset('dogImages/test')

# load list of dog names
dog_names = [item[20:-1] for item in sorted(glob("dogImages/train/*/"))]
# print(dog_names)

# print statistics about the dataset
print('There are %d total dog categories.' % len(dog_names))
print('There are %s total dog images.\n' % len(np.hstack([train_files, valid_files, test_files])))
print('There are %d training dog images.' % len(train_files))
print('There are %d validation dog images.' % len(valid_files))
print('There are %d test dog images.'% len(test_files))
Using TensorFlow backend.
There are 133 total dog categories.
There are 8351 total dog images.

There are 6680 training dog images.
There are 835 validation dog images.
There are 836 test dog images.

Import Human Dataset

In the code cell below, we import a dataset of human images, where the file paths are stored in the numpy array human_files.

In [2]:
import random
random.seed(8675309)

# load filenames in shuffled human dataset
human_files = np.array(glob("lfw/*/*"))
random.shuffle(human_files)

# print statistics about the dataset
print('There are %d total human images.' % len(human_files))
There are 13233 total human images.

Step 1: Detect Humans

We use OpenCV's implementation of Haar feature-based cascade classifiers to detect human faces in images. OpenCV provides many pre-trained face detectors, stored as XML files on github. We have downloaded one of these detectors and stored it in the haarcascades directory.

In the next code cell, we demonstrate how to use this detector to find human faces in a sample image.

In [3]:
import cv2                
import matplotlib.pyplot as plt                        
%matplotlib inline                               

# extract pre-trained face detector
face_cascade = cv2.CascadeClassifier('haarcascades/haarcascade_frontalface_alt.xml')

print(human_files[3])

# load color (BGR) image
img = cv2.imread(human_files[3])
# img = cv2.imread(human_files[10])
# convert BGR image to grayscale
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)

# find faces in image
faces = face_cascade.detectMultiScale(gray)

# print number of faces detected in the image
print('Number of faces detected:', len(faces))

# get bounding box for each detected face
for (x,y,w,h) in faces:
    # add bounding box to color image
    cv2.rectangle(img,(x,y),(x+w,y+h),(255,0,0),2)
    
# convert BGR image to RGB for plotting
cv_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)

# display the image, along with bounding box
plt.imshow(cv_rgb)
plt.show()
lfw/Fernando_Vargas/Fernando_Vargas_0001.jpg
Number of faces detected: 1

Before using any of the face detectors, it is standard procedure to convert the images to grayscale. The detectMultiScale function executes the classifier stored in face_cascade and takes the grayscale image as a parameter.

In the above code, faces is a numpy array of detected faces, where each row corresponds to a detected face. Each detected face is a 1D array with four entries that specifies the bounding box of the detected face. The first two entries in the array (extracted in the above code as x and y) specify the horizontal and vertical positions of the top left corner of the bounding box. The last two entries in the array (extracted here as w and h) specify the width and height of the box.

Write a Human Face Detector

We can use this procedure to write a function that returns True if a human face is detected in an image and False otherwise. This function, aptly named face_detector, takes a string-valued file path to an image as input and appears in the code block below.

In [4]:
# returns "True" if face is detected in image stored at img_path
def face_detector(img_path):
    img = cv2.imread(img_path)
#     print(img_path)
#     display_img(img_path)
    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    faces = face_cascade.detectMultiScale(gray)
    return len(faces) > 0

(IMPLEMENTATION) Assess the Human Face Detector

Question 1: Use the code cell below to test the performance of the face_detector function.

  • What percentage of the first 100 images in human_files have a detected human face?
  • What percentage of the first 100 images in dog_files have a detected human face?

Ideally, we would like 100% of human images with a detected face and 0% of dog images with a detected face. You will see that our algorithm falls short of this goal, but still gives acceptable performance. We extract the file paths for the first 100 images from each of the datasets and store them in the numpy arrays human_files_short and dog_files_short.

Answer:
• 98% of the first 100 images in 'human_files' have a detected human face.
• 11% of the first 100 images in 'dog_files' have a detected human face.

In [5]:
human_files_short = human_files[:100]
dog_files_short = train_files[:100]
# Do NOT modify the code above this line.

## TODO: Test the performance of the face_detector algorithm 
## on the images in human_files_short and dog_files_short. 
def assess_human_detector(filepaths):
    count = 0
    for img_path in filepaths:
        if face_detector(img_path):
            count +=1
    return count

def display_img(img_path):
    img = cv2.imread(img_path)
    cv_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
    plt.imshow(cv_rgb)
    plt.show()



count_human = assess_human_detector(human_files_short)
# Review 1
# human_count = sum([face_detector(human_img) for human_img in human_files_short])

print('Percentage of human images detected as human: {:.1f} %'.format(count_human))
print('False Negative images...')

for img_path in human_files_short:
    if not face_detector(img_path):
        display_img(img_path)

count_dog = assess_human_detector(dog_files_short)
print('Percentage of dog images detected as human: {:.1f} %'.format(count_dog))
print('False Positive images...')

for img_path in dog_files_short:
    if face_detector(img_path):
        display_img(img_path)
Percentage of human images detected as human: 98.0 %
False Negative images...
Percentage of dog images detected as human: 11.0 %
False Positive images...

Question 2: This algorithmic choice necessitates that we communicate to the user that we accept human images only when they provide a clear view of a face (otherwise, we risk having unneccessarily frustrated users!). In your opinion, is this a reasonable expectation to pose on the user? If not, can you think of a way to detect humans in images that does not necessitate an image with a clearly presented face?

Answer:
• For a high precision model, this algorithm is not very good because false positive is quite high, approximately one in ten images, but for a high recall model, this algorithm is by far the best algorithm because false negative is 0% even though it is for the first 100 images. As described above, this app accepts "any user-supplied image as input. If a dog is detected in the image, it will provide an estimate of the dog's breed. If a human is detected, it will provide an estimate of the dog breed that is most resembling." For this app, false positive would not be a big mistake because users eventually want to know an estimate of the dog breed for the image that the user provides. Therefore, this technique for human face detection seems to be adequate.

We suggest the face detector from OpenCV as a potential way to detect human images in your algorithm, but you are free to explore other approaches, especially approaches that make use of deep learning :). Please use the code cell below to design and test your own face detection algorithm. If you decide to pursue this optional task, report performance on each of the datasets.

In [6]:
## (Optional) TODO: Report the performance of another  
## face detection algorithm on the LFW dataset
### Feel free to use as many code cells as needed.

Step 2: Detect Dogs

In this section, we use a pre-trained ResNet-50 model to detect dogs in images. Our first line of code downloads the ResNet-50 model, along with weights that have been trained on ImageNet, a very large, very popular dataset used for image classification and other vision tasks. ImageNet contains over 10 million URLs, each linking to an image containing an object from one of 1000 categories. Given an image, this pre-trained ResNet-50 model returns a prediction (derived from the available categories in ImageNet) for the object that is contained in the image.

In [7]:
from keras.applications.resnet50 import ResNet50

# define ResNet50 model
ResNet50_model = ResNet50(weights='imagenet')

Pre-process the Data

When using TensorFlow as backend, Keras CNNs require a 4D array (which we'll also refer to as a 4D tensor) as input, with shape

$$ (\text{nb_samples}, \text{rows}, \text{columns}, \text{channels}), $$

where nb_samples corresponds to the total number of images (or samples), and rows, columns, and channels correspond to the number of rows, columns, and channels for each image, respectively.

The path_to_tensor function below takes a string-valued file path to a color image as input and returns a 4D tensor suitable for supplying to a Keras CNN. The function first loads the image and resizes it to a square image that is $224 \times 224$ pixels. Next, the image is converted to an array, which is then resized to a 4D tensor. In this case, since we are working with color images, each image has three channels. Likewise, since we are processing a single image (or sample), the returned tensor will always have shape

$$ (1, 224, 224, 3). $$

The paths_to_tensor function takes a numpy array of string-valued image paths as input and returns a 4D tensor with shape

$$ (\text{nb_samples}, 224, 224, 3). $$

Here, nb_samples is the number of samples, or number of images, in the supplied array of image paths. It is best to think of nb_samples as the number of 3D tensors (where each 3D tensor corresponds to a different image) in your dataset!

In [8]:
from keras.preprocessing import image                  
from tqdm import tqdm

def path_to_tensor(img_path):
    # loads RGB image as PIL.Image.Image type
    img = image.load_img(img_path, target_size=(224, 224))
    # convert PIL.Image.Image type to 3D tensor with shape (224, 224, 3)
    x = image.img_to_array(img)
    # convert 3D tensor to 4D tensor with shape (1, 224, 224, 3) and return 4D tensor
    return np.expand_dims(x, axis=0)

def paths_to_tensor(img_paths):
    list_of_tensors = [path_to_tensor(img_path) for img_path in tqdm(img_paths)]
    return np.vstack(list_of_tensors)

Making Predictions with ResNet-50

Getting the 4D tensor ready for ResNet-50, and for any other pre-trained model in Keras, requires some additional processing. First, the RGB image is converted to BGR by reordering the channels. All pre-trained models have the additional normalization step that the mean pixel (expressed in RGB as $[103.939, 116.779, 123.68]$ and calculated from all pixels in all images in ImageNet) must be subtracted from every pixel in each image. This is implemented in the imported function preprocess_input. If you're curious, you can check the code for preprocess_input here.

Now that we have a way to format our image for supplying to ResNet-50, we are now ready to use the model to extract the predictions. This is accomplished with the predict method, which returns an array whose $i$-th entry is the model's predicted probability that the image belongs to the $i$-th ImageNet category. This is implemented in the ResNet50_predict_labels function below.

By taking the argmax of the predicted probability vector, we obtain an integer corresponding to the model's predicted object class, which we can identify with an object category through the use of this dictionary.

In [9]:
from keras.applications.resnet50 import preprocess_input, decode_predictions

def ResNet50_predict_labels(img_path):
    # returns prediction vector for image located at img_path
    img = preprocess_input(path_to_tensor(img_path))
    return np.argmax(ResNet50_model.predict(img))

Write a Dog Detector

While looking at the dictionary, you will notice that the categories corresponding to dogs appear in an uninterrupted sequence and correspond to dictionary keys 151-268, inclusive, to include all categories from 'Chihuahua' to 'Mexican hairless'. Thus, in order to check to see if an image is predicted to contain a dog by the pre-trained ResNet-50 model, we need only check if the ResNet50_predict_labels function above returns a value between 151 and 268 (inclusive).

We use these ideas to complete the dog_detector function below, which returns True if a dog is detected in an image (and False if not).

In [10]:
### returns "True" if a dog is detected in the image stored at img_path
def dog_detector(img_path):
    prediction = ResNet50_predict_labels(img_path)
    return ((prediction <= 268) & (prediction >= 151)) 

(IMPLEMENTATION) Assess the Dog Detector

Question 3: Use the code cell below to test the performance of your dog_detector function.

  • What percentage of the images in human_files_short have a detected dog?
  • What percentage of the images in dog_files_short have a detected dog?

Answer:
• 2% of the first 100 images in 'human_files' have a detected dog.
• 100% of the first 100 images in 'dog_files' have a detected dog.

In [11]:
### TODO: Test the performance of the dog_detector function
### on the images in human_files_short and dog_files_short.
def assess_dog_detector(filepaths):
    count = 0
    for img_path in filepaths:
        if dog_detector(img_path):
            count +=1
    return count

count_human = assess_dog_detector(human_files_short)
print('Percentage of human images detected as dog: {:.1f} %'.format(count_human))
print('False Positive images...')

for img_path in human_files_short:
    if dog_detector(img_path):
        display_img(img_path)

count_dog = assess_dog_detector(dog_files_short)
print('Percentage of dog images detected as dog: {:.1f} %'.format(count_dog))
print('False Negative images...')

for img_path in dog_files_short:
    if not dog_detector(img_path):
        display_img(img_path)
Percentage of human images detected as dog: 2.0 %
False Positive images...
Percentage of dog images detected as dog: 100.0 %
False Negative images...

Step 3: Create a CNN to Classify Dog Breeds (from Scratch)

Now that we have functions for detecting humans and dogs in images, we need a way to predict breed from images. In this step, you will create a CNN that classifies dog breeds. You must create your CNN from scratch (so, you can't use transfer learning yet!), and you must attain a test accuracy of at least 1%. In Step 5 of this notebook, you will have the opportunity to use transfer learning to create a CNN that attains greatly improved accuracy.

Be careful with adding too many trainable layers! More parameters means longer training, which means you are more likely to need a GPU to accelerate the training process. Thankfully, Keras provides a handy estimate of the time that each epoch is likely to take; you can extrapolate this estimate to figure out how long it will take for your algorithm to train.

We mention that the task of assigning breed to dogs from images is considered exceptionally challenging. To see why, consider that even a human would have great difficulty in distinguishing between a Brittany and a Welsh Springer Spaniel.

Brittany Welsh Springer Spaniel

It is not difficult to find other dog breed pairs with minimal inter-class variation (for instance, Curly-Coated Retrievers and American Water Spaniels).

Curly-Coated Retriever American Water Spaniel

Likewise, recall that labradors come in yellow, chocolate, and black. Your vision-based algorithm will have to conquer this high intra-class variation to determine how to classify all of these different shades as the same breed.

Yellow Labrador Chocolate Labrador Black Labrador

We also mention that random chance presents an exceptionally low bar: setting aside the fact that the classes are slightly imabalanced, a random guess will provide a correct answer roughly 1 in 133 times, which corresponds to an accuracy of less than 1%.

Remember that the practice is far ahead of the theory in deep learning. Experiment with many different architectures, and trust your intuition. And, of course, have fun!

Pre-process the Data

We rescale the images by dividing every pixel in every image by 255.

In [12]:
from PIL import ImageFile                            
ImageFile.LOAD_TRUNCATED_IMAGES = True                 

# pre-process the data for Keras
train_tensors = paths_to_tensor(train_files).astype('float32')/255
valid_tensors = paths_to_tensor(valid_files).astype('float32')/255
test_tensors = paths_to_tensor(test_files).astype('float32')/255
100%|██████████| 6680/6680 [00:54<00:00, 122.22it/s]
100%|██████████| 835/835 [00:06<00:00, 136.41it/s]
100%|██████████| 836/836 [00:06<00:00, 136.54it/s]

(IMPLEMENTATION) Model Architecture

Create a CNN to classify dog breed. At the end of your code cell block, summarize the layers of your model by executing the line:

    model.summary()

We have imported some Python modules to get you started, but feel free to import as many modules as you need. If you end up getting stuck, here's a hint that specifies a model that trains relatively fast on CPU and attains >1% test accuracy in 5 epochs:

Sample CNN

Question 4: Outline the steps you took to get to your final CNN architecture and your reasoning at each step. If you chose to use the hinted architecture above, describe why you think that CNN architecture should work well for the image classification task.

Answer: • At first, I added to or removed from the hinted architecture one convolutional layer and max pooling layer, but it did not seem to be working well. Then I changed kernel_size 2 to 3 to see a little bigger part of a image, but it still did not get better. Then I added fully connected layers with dropout layers instead of the global average pooling (GAP) layer because the GAP layer would get rid of some information of images. The resulting accuracy got much better.

In [13]:
from keras.layers import Conv2D, MaxPooling2D, GlobalAveragePooling2D
from keras.layers import Dropout, Flatten, Dense
from keras.models import Sequential

model = Sequential()

### TODO: Define your architecture. 
model.add(Conv2D(filters=16, kernel_size=2, padding='same', activation='relu', input_shape=(224, 224, 3)))
model.add(MaxPooling2D(pool_size=2))
model.add(Conv2D(filters=32, kernel_size=2, padding='same', activation='relu'))
model.add(MaxPooling2D(pool_size=2))
model.add(Conv2D(filters=64, kernel_size=2, padding='same', activation='relu'))
model.add(MaxPooling2D(pool_size=2))
model.add(GlobalAveragePooling2D())
model.add(Dense(133, activation='softmax'))

model.summary()
# Test accuracy: 2.5120%
WARNING:tensorflow:From /usr/local/lib/python3.5/dist-packages/keras/backend/tensorflow_backend.py:1123: calling reduce_mean (from tensorflow.python.ops.math_ops) with keep_dims is deprecated and will be removed in a future version.
Instructions for updating:
keep_dims is deprecated, use keepdims instead
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d_1 (Conv2D)            (None, 224, 224, 16)      208       
_________________________________________________________________
max_pooling2d_2 (MaxPooling2 (None, 112, 112, 16)      0         
_________________________________________________________________
conv2d_2 (Conv2D)            (None, 112, 112, 32)      2080      
_________________________________________________________________
max_pooling2d_3 (MaxPooling2 (None, 56, 56, 32)        0         
_________________________________________________________________
conv2d_3 (Conv2D)            (None, 56, 56, 64)        8256      
_________________________________________________________________
max_pooling2d_4 (MaxPooling2 (None, 28, 28, 64)        0         
_________________________________________________________________
global_average_pooling2d_1 ( (None, 64)                0         
_________________________________________________________________
dense_1 (Dense)              (None, 133)               8645      
=================================================================
Total params: 19,189.0
Trainable params: 19,189.0
Non-trainable params: 0.0
_________________________________________________________________
In [14]:
# Add another layer to the beginning with filters 8. 

model = Sequential()

### TODO: Define your architecture. 
model.add(Conv2D(filters=8, kernel_size=2, padding='same', activation='relu', input_shape=(224, 224, 3)))
model.add(MaxPooling2D(pool_size=2))
model.add(Conv2D(filters=16, kernel_size=2, padding='same', activation='relu'))
model.add(MaxPooling2D(pool_size=2))
model.add(Conv2D(filters=32, kernel_size=2, padding='same', activation='relu'))
model.add(MaxPooling2D(pool_size=2))
model.add(Conv2D(filters=64, kernel_size=2, padding='same', activation='relu'))
model.add(MaxPooling2D(pool_size=2))
model.add(GlobalAveragePooling2D())
model.add(Dense(133, activation='softmax'))

model.summary()
# Test accuracy: 2.3923%
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d_4 (Conv2D)            (None, 224, 224, 8)       104       
_________________________________________________________________
max_pooling2d_5 (MaxPooling2 (None, 112, 112, 8)       0         
_________________________________________________________________
conv2d_5 (Conv2D)            (None, 112, 112, 16)      528       
_________________________________________________________________
max_pooling2d_6 (MaxPooling2 (None, 56, 56, 16)        0         
_________________________________________________________________
conv2d_6 (Conv2D)            (None, 56, 56, 32)        2080      
_________________________________________________________________
max_pooling2d_7 (MaxPooling2 (None, 28, 28, 32)        0         
_________________________________________________________________
conv2d_7 (Conv2D)            (None, 28, 28, 64)        8256      
_________________________________________________________________
max_pooling2d_8 (MaxPooling2 (None, 14, 14, 64)        0         
_________________________________________________________________
global_average_pooling2d_2 ( (None, 64)                0         
_________________________________________________________________
dense_2 (Dense)              (None, 133)               8645      
=================================================================
Total params: 19,613.0
Trainable params: 19,613.0
Non-trainable params: 0.0
_________________________________________________________________
In [15]:
# Remove the layer with filters 16. 

model = Sequential()

### TODO: Define your architecture. 
# model.add(Conv2D(filters=16, kernel_size=2, padding='same', activation='relu'), input_shape=(224, 224, 3))
# model.add(MaxPooling2D(pool_size=2))
model.add(Conv2D(filters=32, kernel_size=2, padding='same', activation='relu', input_shape=(224, 224, 3)))
model.add(MaxPooling2D(pool_size=2))
model.add(Conv2D(filters=64, kernel_size=2, padding='same', activation='relu'))
model.add(MaxPooling2D(pool_size=2))
model.add(GlobalAveragePooling2D())
model.add(Dense(133, activation='softmax'))

model.summary()
#  Test accuracy: 1.6746%
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d_8 (Conv2D)            (None, 224, 224, 32)      416       
_________________________________________________________________
max_pooling2d_9 (MaxPooling2 (None, 112, 112, 32)      0         
_________________________________________________________________
conv2d_9 (Conv2D)            (None, 112, 112, 64)      8256      
_________________________________________________________________
max_pooling2d_10 (MaxPooling (None, 56, 56, 64)        0         
_________________________________________________________________
global_average_pooling2d_3 ( (None, 64)                0         
_________________________________________________________________
dense_3 (Dense)              (None, 133)               8645      
=================================================================
Total params: 17,317.0
Trainable params: 17,317.0
Non-trainable params: 0.0
_________________________________________________________________
In [16]:
# Remove the layer with filters 64. 

model = Sequential()

### TODO: Define your architecture. 
model.add(Conv2D(filters=16, kernel_size=2, padding='same', activation='relu', input_shape=(224, 224, 3)))
model.add(MaxPooling2D(pool_size=2))
model.add(Conv2D(filters=32, kernel_size=2, padding='same', activation='relu'))
model.add(MaxPooling2D(pool_size=2))
# model.add(Conv2D(filters=64, kernel_size=2, padding='same', activation='relu'))
# model.add(MaxPooling2D(pool_size=2))
model.add(GlobalAveragePooling2D())
model.add(Dense(133, activation='softmax'))

model.summary()
# Test accuracy: 1.9139%
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d_10 (Conv2D)           (None, 224, 224, 16)      208       
_________________________________________________________________
max_pooling2d_11 (MaxPooling (None, 112, 112, 16)      0         
_________________________________________________________________
conv2d_11 (Conv2D)           (None, 112, 112, 32)      2080      
_________________________________________________________________
max_pooling2d_12 (MaxPooling (None, 56, 56, 32)        0         
_________________________________________________________________
global_average_pooling2d_4 ( (None, 32)                0         
_________________________________________________________________
dense_4 (Dense)              (None, 133)               4389      
=================================================================
Total params: 6,677.0
Trainable params: 6,677.0
Non-trainable params: 0.0
_________________________________________________________________
In [17]:
# Change kernel_size 2 to 3 for every Convolutional layer.

model = Sequential()

### TODO: Define your architecture. 
model.add(Conv2D(filters=16, kernel_size=3, padding='same', activation='relu', input_shape=(224, 224, 3)))
model.add(MaxPooling2D(pool_size=2))
model.add(Conv2D(filters=32, kernel_size=3, padding='same', activation='relu'))
model.add(MaxPooling2D(pool_size=2))
model.add(Conv2D(filters=64, kernel_size=3, padding='same', activation='relu'))
model.add(MaxPooling2D(pool_size=2))
model.add(GlobalAveragePooling2D())
model.add(Dense(133, activation='softmax'))

model.summary()
# Test accuracy: 2.3923%
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d_12 (Conv2D)           (None, 224, 224, 16)      448       
_________________________________________________________________
max_pooling2d_13 (MaxPooling (None, 112, 112, 16)      0         
_________________________________________________________________
conv2d_13 (Conv2D)           (None, 112, 112, 32)      4640      
_________________________________________________________________
max_pooling2d_14 (MaxPooling (None, 56, 56, 32)        0         
_________________________________________________________________
conv2d_14 (Conv2D)           (None, 56, 56, 64)        18496     
_________________________________________________________________
max_pooling2d_15 (MaxPooling (None, 28, 28, 64)        0         
_________________________________________________________________
global_average_pooling2d_5 ( (None, 64)                0         
_________________________________________________________________
dense_5 (Dense)              (None, 133)               8645      
=================================================================
Total params: 32,229.0
Trainable params: 32,229.0
Non-trainable params: 0.0
_________________________________________________________________
In [18]:
# Add 2 Dense layers with Dropout layers instead of GAP layer. 

model = Sequential()

### TODO: Define your architecture. 
model.add(Conv2D(filters=16, kernel_size=2, padding='same', activation='relu', input_shape=(224, 224, 3)))
model.add(MaxPooling2D(pool_size=2))
model.add(Conv2D(filters=32, kernel_size=2, padding='same', activation='relu'))
model.add(MaxPooling2D(pool_size=2))
model.add(Conv2D(filters=64, kernel_size=2, padding='same', activation='relu'))
model.add(MaxPooling2D(pool_size=2))
# model.add(GlobalAveragePooling2D())
model.add(Dropout(0.3))
model.add(Flatten())
model.add(Dense(500, activation='relu'))
model.add(Dropout(0.4))
model.add(Dense(133, activation='softmax'))

model.summary()
# Test accuracy: 9.3301%
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d_15 (Conv2D)           (None, 224, 224, 16)      208       
_________________________________________________________________
max_pooling2d_16 (MaxPooling (None, 112, 112, 16)      0         
_________________________________________________________________
conv2d_16 (Conv2D)           (None, 112, 112, 32)      2080      
_________________________________________________________________
max_pooling2d_17 (MaxPooling (None, 56, 56, 32)        0         
_________________________________________________________________
conv2d_17 (Conv2D)           (None, 56, 56, 64)        8256      
_________________________________________________________________
max_pooling2d_18 (MaxPooling (None, 28, 28, 64)        0         
_________________________________________________________________
dropout_1 (Dropout)          (None, 28, 28, 64)        0         
_________________________________________________________________
flatten_2 (Flatten)          (None, 50176)             0         
_________________________________________________________________
dense_6 (Dense)              (None, 500)               25088500  
_________________________________________________________________
dropout_2 (Dropout)          (None, 500)               0         
_________________________________________________________________
dense_7 (Dense)              (None, 133)               66633     
=================================================================
Total params: 25,165,677.0
Trainable params: 25,165,677.0
Non-trainable params: 0.0
_________________________________________________________________

Compile the Model

In [19]:
model.compile(optimizer='rmsprop', loss='categorical_crossentropy', metrics=['accuracy'])
WARNING:tensorflow:From /usr/local/lib/python3.5/dist-packages/keras/backend/tensorflow_backend.py:2548: calling reduce_sum (from tensorflow.python.ops.math_ops) with keep_dims is deprecated and will be removed in a future version.
Instructions for updating:
keep_dims is deprecated, use keepdims instead

(IMPLEMENTATION) Train the Model

Train your model in the code cell below. Use model checkpointing to save the model that attains the best validation loss.

You are welcome to augment the training data, but this is not a requirement.

In [20]:
from keras.callbacks import ModelCheckpoint  

### TODO: specify the number of epochs that you would like to use to train the model.

epochs = 5

### Do NOT modify the code below this line.

checkpointer = ModelCheckpoint(filepath='saved_models/weights.best.from_scratch.hdf5', 
                               verbose=1, save_best_only=True)

model.fit(train_tensors, train_targets, 
          validation_data=(valid_tensors, valid_targets),
          epochs=epochs, batch_size=20, callbacks=[checkpointer], verbose=1)         

# Review 1, you can use Early Stopping. 
# from keras.callbacks import EarlyStopping
# e_stop=EarlyStopping(monitor='val_acc', patience=10, verbose=2, mode='auto')
# model.fit(train_Xception, train_targets, 
#               validation_data=(valid_Xception, valid_targets),
#               epochs=32, batch_size=32, callbacks=[checkpointer,e_stop], verbose=1)
Train on 6680 samples, validate on 835 samples
Epoch 1/5
6660/6680 [============================>.] - ETA: 0s - loss: 4.8921 - acc: 0.0162Epoch 00000: val_loss improved from inf to 4.64840, saving model to saved_models/weights.best.from_scratch.hdf5
6680/6680 [==============================] - 24s - loss: 4.8917 - acc: 0.0162 - val_loss: 4.6484 - val_acc: 0.0431
Epoch 2/5
6660/6680 [============================>.] - ETA: 0s - loss: 4.4596 - acc: 0.0524Epoch 00001: val_loss improved from 4.64840 to 4.25214, saving model to saved_models/weights.best.from_scratch.hdf5
6680/6680 [==============================] - 23s - loss: 4.4580 - acc: 0.0525 - val_loss: 4.2521 - val_acc: 0.0635
Epoch 3/5
6660/6680 [============================>.] - ETA: 0s - loss: 3.9291 - acc: 0.1140Epoch 00002: val_loss improved from 4.25214 to 4.15442, saving model to saved_models/weights.best.from_scratch.hdf5
6680/6680 [==============================] - 22s - loss: 3.9289 - acc: 0.1142 - val_loss: 4.1544 - val_acc: 0.0862
Epoch 4/5
6660/6680 [============================>.] - ETA: 0s - loss: 3.1051 - acc: 0.2661Epoch 00003: val_loss did not improve
6680/6680 [==============================] - 22s - loss: 3.1048 - acc: 0.2662 - val_loss: 4.2361 - val_acc: 0.0982
Epoch 5/5
6660/6680 [============================>.] - ETA: 0s - loss: 2.0769 - acc: 0.4851Epoch 00004: val_loss did not improve
6680/6680 [==============================] - 22s - loss: 2.0790 - acc: 0.4847 - val_loss: 4.6831 - val_acc: 0.0994
Out[20]:
<keras.callbacks.History at 0x7f253f265828>

Load the Model with the Best Validation Loss

In [21]:
model.load_weights('saved_models/weights.best.from_scratch.hdf5')

Test the Model

Try out your model on the test dataset of dog images. Ensure that your test accuracy is greater than 1%.

In [22]:
# get index of predicted dog breed for each image in test set
dog_breed_predictions = [np.argmax(model.predict(np.expand_dims(tensor, axis=0))) for tensor in test_tensors]

# report test accuracy
test_accuracy = 100*np.sum(np.array(dog_breed_predictions)==np.argmax(test_targets, axis=1))/len(dog_breed_predictions)
print('Test accuracy: %.4f%%' % test_accuracy)
Test accuracy: 9.3301%

Step 4: Use a CNN to Classify Dog Breeds

To reduce training time without sacrificing accuracy, we show you how to train a CNN using transfer learning. In the following step, you will get a chance to use transfer learning to train your own CNN.

Obtain Bottleneck Features

In [89]:
bottleneck_features = np.load('bottleneck_features/DogVGG16Data.npz')
train_VGG16 = bottleneck_features['train']
valid_VGG16 = bottleneck_features['valid']
test_VGG16 = bottleneck_features['test']

Model Architecture

The model uses the the pre-trained VGG-16 model as a fixed feature extractor, where the last convolutional output of VGG-16 is fed as input to our model. We only add a global average pooling layer and a fully connected layer, where the latter contains one node for each dog category and is equipped with a softmax.

In [90]:
VGG16_model = Sequential()
VGG16_model.add(GlobalAveragePooling2D(input_shape=train_VGG16.shape[1:]))
VGG16_model.add(Dense(133, activation='softmax'))

VGG16_model.summary()
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
global_average_pooling2d_17  (None, 512)               0         
_________________________________________________________________
dense_21 (Dense)             (None, 133)               68229     
=================================================================
Total params: 68,229.0
Trainable params: 68,229.0
Non-trainable params: 0.0
_________________________________________________________________

Compile the Model

In [91]:
VGG16_model.compile(loss='categorical_crossentropy', optimizer='rmsprop', metrics=['accuracy'])

Train the Model

In [92]:
checkpointer = ModelCheckpoint(filepath='saved_models/weights.best.VGG16.hdf5', 
                               verbose=1, save_best_only=True)

VGG16_model.fit(train_VGG16, train_targets, 
          validation_data=(valid_VGG16, valid_targets),
          epochs=20, batch_size=20, callbacks=[checkpointer], verbose=1)
Train on 6680 samples, validate on 835 samples
Epoch 1/20
6620/6680 [============================>.] - ETA: 0s - loss: 12.4078 - acc: 0.1169Epoch 00000: val_loss improved from inf to 11.13725, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 4s - loss: 12.3979 - acc: 0.1174 - val_loss: 11.1372 - val_acc: 0.2012
Epoch 2/20
6580/6680 [============================>.] - ETA: 0s - loss: 10.2305 - acc: 0.2749Epoch 00001: val_loss improved from 11.13725 to 10.13245, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s - loss: 10.2262 - acc: 0.2757 - val_loss: 10.1325 - val_acc: 0.2802
Epoch 3/20
6580/6680 [============================>.] - ETA: 0s - loss: 9.6138 - acc: 0.3403Epoch 00002: val_loss improved from 10.13245 to 9.72639, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s - loss: 9.6033 - acc: 0.3412 - val_loss: 9.7264 - val_acc: 0.3186
Epoch 4/20
6640/6680 [============================>.] - ETA: 0s - loss: 9.3000 - acc: 0.3779Epoch 00003: val_loss improved from 9.72639 to 9.59918, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s - loss: 9.3023 - acc: 0.3780 - val_loss: 9.5992 - val_acc: 0.3293
Epoch 5/20
6620/6680 [============================>.] - ETA: 0s - loss: 9.0654 - acc: 0.4009Epoch 00004: val_loss improved from 9.59918 to 9.46251, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s - loss: 9.0692 - acc: 0.4007 - val_loss: 9.4625 - val_acc: 0.3365
Epoch 6/20
6640/6680 [============================>.] - ETA: 0s - loss: 8.9837 - acc: 0.4185Epoch 00005: val_loss did not improve
6680/6680 [==============================] - 2s - loss: 8.9860 - acc: 0.4183 - val_loss: 9.5002 - val_acc: 0.3425
Epoch 7/20
6540/6680 [============================>.] - ETA: 0s - loss: 8.9807 - acc: 0.4240Epoch 00006: val_loss did not improve
6680/6680 [==============================] - 2s - loss: 8.9481 - acc: 0.4260 - val_loss: 9.5514 - val_acc: 0.3401
Epoch 8/20
6500/6680 [============================>.] - ETA: 0s - loss: 8.9509 - acc: 0.4292Epoch 00007: val_loss improved from 9.46251 to 9.37888, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s - loss: 8.9194 - acc: 0.4310 - val_loss: 9.3789 - val_acc: 0.3509
Epoch 9/20
6640/6680 [============================>.] - ETA: 0s - loss: 8.8329 - acc: 0.4346Epoch 00008: val_loss improved from 9.37888 to 9.27154, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s - loss: 8.8242 - acc: 0.4352 - val_loss: 9.2715 - val_acc: 0.3545
Epoch 10/20
6580/6680 [============================>.] - ETA: 0s - loss: 8.5634 - acc: 0.4497Epoch 00009: val_loss improved from 9.27154 to 9.16715, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s - loss: 8.5800 - acc: 0.4490 - val_loss: 9.1671 - val_acc: 0.3617
Epoch 11/20
6600/6680 [============================>.] - ETA: 0s - loss: 8.4950 - acc: 0.4585Epoch 00010: val_loss improved from 9.16715 to 9.05192, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s - loss: 8.5034 - acc: 0.4579 - val_loss: 9.0519 - val_acc: 0.3677
Epoch 12/20
6560/6680 [============================>.] - ETA: 0s - loss: 8.3710 - acc: 0.4622Epoch 00011: val_loss improved from 9.05192 to 9.02285, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s - loss: 8.3708 - acc: 0.4623 - val_loss: 9.0228 - val_acc: 0.3749
Epoch 13/20
6640/6680 [============================>.] - ETA: 0s - loss: 8.2169 - acc: 0.4755Epoch 00012: val_loss improved from 9.02285 to 8.90646, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s - loss: 8.2256 - acc: 0.4750 - val_loss: 8.9065 - val_acc: 0.3689
Epoch 14/20
6560/6680 [============================>.] - ETA: 0s - loss: 8.0823 - acc: 0.4829Epoch 00013: val_loss improved from 8.90646 to 8.84385, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s - loss: 8.0774 - acc: 0.4831 - val_loss: 8.8438 - val_acc: 0.3832
Epoch 15/20
6540/6680 [============================>.] - ETA: 0s - loss: 7.9745 - acc: 0.4888Epoch 00014: val_loss improved from 8.84385 to 8.61988, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s - loss: 7.9871 - acc: 0.4874 - val_loss: 8.6199 - val_acc: 0.3940
Epoch 16/20
6600/6680 [============================>.] - ETA: 0s - loss: 7.8716 - acc: 0.4991Epoch 00015: val_loss did not improve
6680/6680 [==============================] - 2s - loss: 7.8655 - acc: 0.4994 - val_loss: 8.7262 - val_acc: 0.3856
Epoch 17/20
6580/6680 [============================>.] - ETA: 0s - loss: 7.8105 - acc: 0.5067Epoch 00016: val_loss improved from 8.61988 to 8.55643, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s - loss: 7.8198 - acc: 0.5061 - val_loss: 8.5564 - val_acc: 0.4012
Epoch 18/20
6520/6680 [============================>.] - ETA: 0s - loss: 7.7767 - acc: 0.5080Epoch 00017: val_loss did not improve
6680/6680 [==============================] - 2s - loss: 7.7480 - acc: 0.5097 - val_loss: 8.5620 - val_acc: 0.4012
Epoch 19/20
6660/6680 [============================>.] - ETA: 0s - loss: 7.6031 - acc: 0.5135Epoch 00018: val_loss improved from 8.55643 to 8.33592, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s - loss: 7.5949 - acc: 0.5141 - val_loss: 8.3359 - val_acc: 0.4144
Epoch 20/20
6540/6680 [============================>.] - ETA: 0s - loss: 7.4120 - acc: 0.5283Epoch 00019: val_loss improved from 8.33592 to 8.27083, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s - loss: 7.4221 - acc: 0.5275 - val_loss: 8.2708 - val_acc: 0.4168
Out[92]:
<keras.callbacks.History at 0x7fdb60ae4710>

Load the Model with the Best Validation Loss

In [93]:
VGG16_model.load_weights('saved_models/weights.best.VGG16.hdf5')

Test the Model

Now, we can use the CNN to test how well it identifies breed within our test dataset of dog images. We print the test accuracy below.

In [94]:
# get index of predicted dog breed for each image in test set
VGG16_predictions = [np.argmax(VGG16_model.predict(np.expand_dims(feature, axis=0))) for feature in test_VGG16]

# report test accuracy
test_accuracy = 100*np.sum(np.array(VGG16_predictions)==np.argmax(test_targets, axis=1))/len(VGG16_predictions)
print('Test accuracy: %.4f%%' % test_accuracy)
Test accuracy: 43.0622%

Predict Dog Breed with the Model

In [47]:
from extract_bottleneck_features import *

# def VGG16_predict_breed(img_path):
#     # extract bottleneck features
#     bottleneck_feature = extract_VGG16(path_to_tensor(img_path))
#     # obtain predicted vector
#     predicted_vector = VGG16_model.predict(bottleneck_feature)
#     # return dog breed that is predicted by the model
#     return dog_names[np.argmax(predicted_vector)]

Step 5: Create a CNN to Classify Dog Breeds (using Transfer Learning)

You will now use transfer learning to create a CNN that can identify dog breed from images. Your CNN must attain at least 60% accuracy on the test set.

In Step 4, we used transfer learning to create a CNN using VGG-16 bottleneck features. In this section, you must use the bottleneck features from a different pre-trained model. To make things easier for you, we have pre-computed the features for all of the networks that are currently available in Keras:

The files are encoded as such:

Dog{network}Data.npz

where {network}, in the above filename, can be one of VGG19, Resnet50, InceptionV3, or Xception. Pick one of the above architectures, download the corresponding bottleneck features, and store the downloaded file in the bottleneck_features/ folder in the repository.

(IMPLEMENTATION) Obtain Bottleneck Features

In the code block below, extract the bottleneck features corresponding to the train, test, and validation sets by running the following:

bottleneck_features = np.load('bottleneck_features/Dog{network}Data.npz')
train_{network} = bottleneck_features['train']
valid_{network} = bottleneck_features['valid']
test_{network} = bottleneck_features['test']
In [13]:
### TODO: Obtain bottleneck features from another pre-trained CNN.
# bottleneck_features = np.load('bottleneck_features/DogVGG19Data.npz')
# train_VGG19 = bottleneck_features['train']
# valid_VGG19 = bottleneck_features['valid']
# test_VGG19 = bottleneck_features['test']
bottleneck_features = np.load('bottleneck_features/DogXceptionData.npz')
train_Xception = bottleneck_features['train']
valid_Xception = bottleneck_features['valid']
test_Xception = bottleneck_features['test']

(IMPLEMENTATION) Model Architecture

Create a CNN to classify dog breed. At the end of your code cell block, summarize the layers of your model by executing the line:

    <your model's name>.summary()

Question 5: Outline the steps you took to get to your final CNN architecture and your reasoning at each step. Describe why you think the architecture is suitable for the current problem.

Answer: • Firstly, I tried VGG19 model, but I was tempted to use Xception because it is the newest of the four models. At first, I took a look at the accuracy for the same additional layers as the model with the VGG19. The model with Xception is much better than the model with VGG19. Then, I added one more fully connected layer and Dropout because adding these layers improved the accuracy of the model in the previous section.

In [102]:
### TODO: Define your architecture.
VGG19_model = Sequential()
VGG19_model.add(GlobalAveragePooling2D(input_shape=train_VGG19.shape[1:]))
VGG19_model.add(Dense(133, activation='softmax'))

VGG19_model.summary()
#  Test accuracy: 46.2919%
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
global_average_pooling2d_19  (None, 512)               0         
_________________________________________________________________
dense_23 (Dense)             (None, 133)               68229     
=================================================================
Total params: 68,229.0
Trainable params: 68,229.0
Non-trainable params: 0.0
_________________________________________________________________
In [24]:
# Use Xception architecture. 

Xception_model = Sequential()
Xception_model.add(GlobalAveragePooling2D(input_shape=train_Xception.shape[1:]))
Xception_model.add(Dense(133, activation='softmax'))

Xception_model.summary()
#  Test accuracy: 83.4928%
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
global_average_pooling2d_6 ( (None, 2048)              0         
_________________________________________________________________
dense_8 (Dense)              (None, 133)               272517    
=================================================================
Total params: 272,517.0
Trainable params: 272,517.0
Non-trainable params: 0.0
_________________________________________________________________
In [14]:
# Add 2 Dense layers with Dropout layers instead of GAP layer. 
from keras.layers import Conv2D, MaxPooling2D, GlobalAveragePooling2D
from keras.layers import Dropout, Flatten, Dense
from keras.models import Sequential

Xception_model = Sequential()
Xception_model.add(GlobalAveragePooling2D(input_shape=train_Xception.shape[1:]))
Xception_model.add(Dropout(0.3))
Xception_model.add(Dense(500, activation='relu'))
Xception_model.add(Dropout(0.4))
Xception_model.add(Dense(133, activation='softmax'))

Xception_model.summary()
#  Test accuracy: 82.1770%
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
global_average_pooling2d_1 ( (None, 2048)              0         
_________________________________________________________________
dropout_1 (Dropout)          (None, 2048)              0         
_________________________________________________________________
dense_1 (Dense)              (None, 500)               1024500   
_________________________________________________________________
dropout_2 (Dropout)          (None, 500)               0         
_________________________________________________________________
dense_2 (Dense)              (None, 133)               66633     
=================================================================
Total params: 1,091,133.0
Trainable params: 1,091,133.0
Non-trainable params: 0.0
_________________________________________________________________
In [15]:
# Remove a Dropout layer.
Xception_model = Sequential()
Xception_model.add(GlobalAveragePooling2D(input_shape=train_Xception.shape[1:]))
Xception_model.add(Dense(500, activation='relu'))
Xception_model.add(Dropout(0.5))
Xception_model.add(Dense(133, activation='softmax'))

Xception_model.summary()
#  Test accuracy: 85.0478% (rmsprop)
#  Test accuracy: 84.0909% (SGD, lr=0.01, momentum=0.9, decay=1e-6)
#  Test accuracy: 84.9282% (SGD, lr=0.01, momentum=0.0, decay=0.0)
#  Test accuracy: 83.1340% (Adam)
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
global_average_pooling2d_2 ( (None, 2048)              0         
_________________________________________________________________
dense_3 (Dense)              (None, 500)               1024500   
_________________________________________________________________
dropout_3 (Dropout)          (None, 500)               0         
_________________________________________________________________
dense_4 (Dense)              (None, 133)               66633     
=================================================================
Total params: 1,091,133.0
Trainable params: 1,091,133.0
Non-trainable params: 0.0
_________________________________________________________________

(IMPLEMENTATION) Compile the Model

In [16]:
### TODO: Compile the model.
# Xception_model.compile(loss='categorical_crossentropy', optimizer='rmsprop', metrics=['accuracy'])

# Review 2 
from keras import optimizers
# sgd= optimizers.SGD(lr=0.01, momentum=0.0, decay=0.0)
# sgd= optimizers.SGD(lr=0.01, momentum=0.9, decay=1e-6)
# Xception_model.compile(loss='categorical_crossentropy', optimizer=sgd, metrics=['accuracy'])
Xception_model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])

(IMPLEMENTATION) Train the Model

Train your model in the code cell below. Use model checkpointing to save the model that attains the best validation loss.

You are welcome to augment the training data, but this is not a requirement.

In [17]:
### TODO: Train the model.
from keras.callbacks import ModelCheckpoint  

checkpointer = ModelCheckpoint(filepath='saved_models/weights.best.Xception.hdf5', 
                               verbose=1, save_best_only=True)

# Xception_model.fit(train_Xception, train_targets, 
#           validation_data=(valid_Xception, valid_targets),
#           epochs=20, batch_size=20, callbacks=[checkpointer], verbose=1)

# Early stopping.
from keras.callbacks import EarlyStopping
e_stop=EarlyStopping(monitor='val_loss', patience=15, verbose=2)

model_history = Xception_model.fit(train_Xception, train_targets, 
              validation_data=(valid_Xception, valid_targets),
              epochs=50, batch_size=20, callbacks=[checkpointer, e_stop], verbose=1)
Train on 6680 samples, validate on 835 samples
Epoch 1/50
6640/6680 [============================>.] - ETA: 0s - loss: 1.5854 - acc: 0.6191Epoch 00000: val_loss improved from inf to 0.61787, saving model to saved_models/weights.best.Xception.hdf5
6680/6680 [==============================] - 5s - loss: 1.5811 - acc: 0.6199 - val_loss: 0.6179 - val_acc: 0.8036
Epoch 2/50
6620/6680 [============================>.] - ETA: 0s - loss: 0.6918 - acc: 0.7927Epoch 00001: val_loss improved from 0.61787 to 0.57695, saving model to saved_models/weights.best.Xception.hdf5
6680/6680 [==============================] - 4s - loss: 0.6915 - acc: 0.7925 - val_loss: 0.5770 - val_acc: 0.8132
Epoch 3/50
6660/6680 [============================>.] - ETA: 0s - loss: 0.5583 - acc: 0.8300Epoch 00002: val_loss did not improve
6680/6680 [==============================] - 4s - loss: 0.5587 - acc: 0.8299 - val_loss: 0.5948 - val_acc: 0.8251
Epoch 4/50
6620/6680 [============================>.] - ETA: 0s - loss: 0.4840 - acc: 0.8508Epoch 00003: val_loss improved from 0.57695 to 0.56571, saving model to saved_models/weights.best.Xception.hdf5
6680/6680 [==============================] - 4s - loss: 0.4829 - acc: 0.8512 - val_loss: 0.5657 - val_acc: 0.8204
Epoch 5/50
6660/6680 [============================>.] - ETA: 0s - loss: 0.4137 - acc: 0.8671Epoch 00004: val_loss did not improve
6680/6680 [==============================] - 4s - loss: 0.4133 - acc: 0.8672 - val_loss: 0.5849 - val_acc: 0.8144
Epoch 6/50
6640/6680 [============================>.] - ETA: 0s - loss: 0.3723 - acc: 0.8800Epoch 00005: val_loss did not improve
6680/6680 [==============================] - 4s - loss: 0.3736 - acc: 0.8799 - val_loss: 0.5808 - val_acc: 0.8263
Epoch 7/50
6600/6680 [============================>.] - ETA: 0s - loss: 0.3492 - acc: 0.8847Epoch 00006: val_loss improved from 0.56571 to 0.54170, saving model to saved_models/weights.best.Xception.hdf5
6680/6680 [==============================] - 4s - loss: 0.3516 - acc: 0.8840 - val_loss: 0.5417 - val_acc: 0.8359
Epoch 8/50
6620/6680 [============================>.] - ETA: 0s - loss: 0.3025 - acc: 0.9000Epoch 00007: val_loss did not improve
6680/6680 [==============================] - 4s - loss: 0.3034 - acc: 0.8999 - val_loss: 0.6070 - val_acc: 0.8263
Epoch 9/50
6620/6680 [============================>.] - ETA: 0s - loss: 0.3070 - acc: 0.8971Epoch 00008: val_loss did not improve
6680/6680 [==============================] - 4s - loss: 0.3080 - acc: 0.8966 - val_loss: 0.6024 - val_acc: 0.8359
Epoch 10/50
6660/6680 [============================>.] - ETA: 0s - loss: 0.2662 - acc: 0.9078Epoch 00009: val_loss did not improve
6680/6680 [==============================] - 4s - loss: 0.2664 - acc: 0.9076 - val_loss: 0.6147 - val_acc: 0.8275
Epoch 11/50
6640/6680 [============================>.] - ETA: 0s - loss: 0.2534 - acc: 0.9194Epoch 00010: val_loss did not improve
6680/6680 [==============================] - 4s - loss: 0.2537 - acc: 0.9195 - val_loss: 0.6335 - val_acc: 0.8335
Epoch 12/50
6620/6680 [============================>.] - ETA: 0s - loss: 0.2668 - acc: 0.9080Epoch 00011: val_loss did not improve
6680/6680 [==============================] - 4s - loss: 0.2670 - acc: 0.9076 - val_loss: 0.6298 - val_acc: 0.8240
Epoch 13/50
6620/6680 [============================>.] - ETA: 0s - loss: 0.2243 - acc: 0.9272Epoch 00012: val_loss did not improve
6680/6680 [==============================] - 4s - loss: 0.2235 - acc: 0.9275 - val_loss: 0.6556 - val_acc: 0.8359
Epoch 14/50
6660/6680 [============================>.] - ETA: 0s - loss: 0.2066 - acc: 0.9312Epoch 00013: val_loss did not improve
6680/6680 [==============================] - 4s - loss: 0.2060 - acc: 0.9314 - val_loss: 0.6983 - val_acc: 0.8347
Epoch 15/50
6600/6680 [============================>.] - ETA: 0s - loss: 0.1986 - acc: 0.9342Epoch 00014: val_loss did not improve
6680/6680 [==============================] - 4s - loss: 0.1990 - acc: 0.9341 - val_loss: 0.7150 - val_acc: 0.8395
Epoch 16/50
6640/6680 [============================>.] - ETA: 0s - loss: 0.1806 - acc: 0.9405Epoch 00015: val_loss did not improve
6680/6680 [==============================] - 4s - loss: 0.1810 - acc: 0.9406 - val_loss: 0.6706 - val_acc: 0.8383
Epoch 17/50
6600/6680 [============================>.] - ETA: 0s - loss: 0.1756 - acc: 0.9386Epoch 00016: val_loss did not improve
6680/6680 [==============================] - 4s - loss: 0.1764 - acc: 0.9385 - val_loss: 0.7046 - val_acc: 0.8395
Epoch 18/50
6640/6680 [============================>.] - ETA: 0s - loss: 0.1768 - acc: 0.9432Epoch 00017: val_loss did not improve
6680/6680 [==============================] - 4s - loss: 0.1762 - acc: 0.9434 - val_loss: 0.7318 - val_acc: 0.8395
Epoch 19/50
6600/6680 [============================>.] - ETA: 0s - loss: 0.1708 - acc: 0.9453Epoch 00018: val_loss did not improve
6680/6680 [==============================] - 4s - loss: 0.1707 - acc: 0.9452 - val_loss: 0.7108 - val_acc: 0.8431
Epoch 20/50
6660/6680 [============================>.] - ETA: 0s - loss: 0.1724 - acc: 0.9428Epoch 00019: val_loss did not improve
6680/6680 [==============================] - 4s - loss: 0.1728 - acc: 0.9427 - val_loss: 0.7262 - val_acc: 0.8383
Epoch 21/50
6640/6680 [============================>.] - ETA: 0s - loss: 0.1734 - acc: 0.9438Epoch 00020: val_loss did not improve
6680/6680 [==============================] - 4s - loss: 0.1747 - acc: 0.9434 - val_loss: 0.7857 - val_acc: 0.8323
Epoch 22/50
6620/6680 [============================>.] - ETA: 0s - loss: 0.1493 - acc: 0.9518Epoch 00021: val_loss did not improve
6680/6680 [==============================] - 4s - loss: 0.1484 - acc: 0.9521 - val_loss: 0.7659 - val_acc: 0.8407
Epoch 23/50
6640/6680 [============================>.] - ETA: 0s - loss: 0.1641 - acc: 0.9434Epoch 00022: val_loss did not improve
6680/6680 [==============================] - 4s - loss: 0.1637 - acc: 0.9436 - val_loss: 0.7747 - val_acc: 0.8371
Epoch 00022: early stopping
In [18]:
# Display the process of learning. 
plt.plot(model_history.history['loss'])
plt.plot(model_history.history['val_loss'])
plt.title('model history: Loss')
plt.ylabel('loss')
plt.xlabel('epoch')
plt.legend(['loss', 'val_loss'], loc='center left')
plt.show()

plt.plot(model_history.history['acc'])
plt.plot(model_history.history['val_acc'])
plt.title('model history: Accuracy')
plt.ylabel('accuracy')
plt.xlabel('epoch')
plt.legend(['acc', 'val_acc'], loc='center left')
plt.show()

(IMPLEMENTATION) Load the Model with the Best Validation Loss

In [19]:
### TODO: Load the model weights with the best validation loss. 
Xception_model.load_weights('saved_models/weights.best.Xception.hdf5')

(IMPLEMENTATION) Test the Model

Try out your model on the test dataset of dog images. Ensure that your test accuracy is greater than 60%.

In [20]:
### TODO: Calculate classification accuracy on the test dataset. 
# get index of predicted dog breed for each image in test set
Xception_predictions = [np.argmax(Xception_model.predict(np.expand_dims(feature, axis=0))) for feature in test_Xception]

# report test accuracy
test_accuracy = 100*np.sum(np.array(Xception_predictions)==np.argmax(test_targets, axis=1))/len(Xception_predictions)
print('Test accuracy: %.4f%%' % test_accuracy)
Test accuracy: 83.1340%

(IMPLEMENTATION) Predict Dog Breed with the Model

Write a function that takes an image path as input and returns the dog breed (Affenpinscher, Afghan_hound, etc) that is predicted by your model.

Similar to the analogous function in Step 5, your function should have three steps:

  1. Extract the bottleneck features corresponding to the chosen CNN model.
  2. Supply the bottleneck features as input to the model to return the predicted vector. Note that the argmax of this prediction vector gives the index of the predicted dog breed.
  3. Use the dog_names array defined in Step 0 of this notebook to return the corresponding breed.

The functions to extract the bottleneck features can be found in extract_bottleneck_features.py, and they have been imported in an earlier code cell. To obtain the bottleneck features corresponding to your chosen CNN architecture, you need to use the function

extract_{network}

where {network}, in the above filename, should be one of VGG19, Resnet50, InceptionV3, or Xception.

In [21]:
### TODO: Write a function that takes a path to an image as input
### and returns the dog breed that is predicted by the model. 
from extract_bottleneck_features import *

def Xception_predict_breed(img_path):
    # extract bottleneck features
    bottleneck_feature = extract_Xception(path_to_tensor(img_path))
    # obtain predicted vector
    predicted_vector = Xception_model.predict(bottleneck_feature)
    # return dog breed that is predicted by the model
    return dog_names[np.argmax(predicted_vector)]

Step 6: Write your Algorithm

Write an algorithm that accepts a file path to an image and first determines whether the image contains a human, dog, or neither. Then,

  • if a dog is detected in the image, return the predicted breed.
  • if a human is detected in the image, return the resembling dog breed.
  • if neither is detected in the image, provide output that indicates an error.

You are welcome to write your own functions for detecting humans and dogs in images, but feel free to use the face_detector and dog_detector functions developed above. You are required to use your CNN from Step 5 to predict dog breed.

Some sample output for our algorithm is provided below, but feel free to design your own user experience!

Sample Human Output

(IMPLEMENTATION) Write your Algorithm

In [22]:
### TODO: Write your algorithm.
### Feel free to use as many code cells as needed.

def dog_breed_classifier(img_paths):
    for img_path in img_paths:
        # Review 1. dog_detector() should be first because the accuracy of dog detector is higher.
        if dog_detector(img_path): 
            print("hello, dog!")
            display_img(img_path)
            print("Your predicted breed is ...")
            ans = Xception_predict_breed(img_path)
            print(ans)
#             print(glob.glob("dogImages/train/*." + ans + ".jpg"))
#             list = np.array(glob("dogImages/train/*." + ans + ".jpg"))
#             print(list)
#             display_img(list[0])            
            print("")
        elif face_detector(img_path):
            print("hello, human!")
            display_img(img_path)
            print("You look like a ...")
            ans = Xception_predict_breed(img_path)
            print(ans)
#             p = np.array(glob("dogImages/train/*."))
#             print(p)
#             list = np.array(glob("dogImages/train/*." + ans + ".*"))
#             print(list)
#             display_img(list[0])
            print("")  
        else:
            print("hello, something other than human or dog!")
            print("Error: please provide an image with clear view of human face or dog")
            display_img(img_path)
            print("")

Step 7: Test Your Algorithm

In this section, you will take your new algorithm for a spin! What kind of dog does the algorithm think that you look like? If you have a dog, does it predict your dog's breed accurately? If you have a cat, does it mistakenly think that your cat is a dog?

(IMPLEMENTATION) Test Your Algorithm on Sample Images!

Test your algorithm at least six images on your computer. Feel free to use any images you like. Use at least two human and two dog images.

Question 6: Is the output better than you expected :) ? Or worse :( ? Provide at least three possible points of improvement for your algorithm.

Answer:
• I was surprised that this algorithm can classify all of the different color Labradors as the same breed, and distinguish Brittany and Welsh Springer Spaniel.
• For the fifth image, this algorithm's prediction is quite good despite the fact that the object is relatively small at the bottom right of the image.
• As to predicting the resembling breed of the human images, this algoritm does not seem to be working very well.

UPDATE:
• Data Augmentation:
Adding to the training set new images created by doing random rotations or translations, or flipping the images of the training set can avoid overfitting and improve the performance of the algorithm.
• Optimizer:
Changing optimizer might lead to a better performance of the algorithm.
• Ensembling:
Combining with different models which are trained in different ways, such as a model that has a different neural network architecture can improve the performance of the algorithm.

In [23]:
## TODO: Execute your algorithm from Step 6 on
## at least 6 images on your computer.
## Feel free to use as many code cells as needed.

sample_files = np.array(glob("sample_images_for_step7/*"))
dog_breed_classifier(sample_files)

# testfile = ["/home/ubuntu/dog-project/sample_images_for_step7/welsh_springer_spaniel.jpeg"]
# testfile = ["/home/ubuntu/dog-project/sample_images_for_step7/mozart.jpeg"]
# dog_breed_classifier(testfile)
hello, dog!
Your predicted breed is ...
Labrador_retriever

hello, human!
You look like a ...
Norwegian_lundehund

hello, dog!
Your predicted breed is ...
Brittany

hello, dog!
Your predicted breed is ...
Flat-coated_retriever

hello, something other than human or dog!
Error: please provide an image with clear view of human face or dog
hello, dog!
Your predicted breed is ...
Beagle

hello, human!
You look like a ...
Cardigan_welsh_corgi

hello, dog!
Your predicted breed is ...
Bedlington_terrier

hello, dog!
Your predicted breed is ...
Labrador_retriever

hello, dog!
Your predicted breed is ...
Labrador_retriever

In [ ]: